KAT6 inhibition as a candidate aging intervention

KAT6A and KAT6B belong to the MYST family, the histone acetyltransferases numbered KAT5 to KAT8, and the two of them assemble into the same four-protein complex and write the same histone mark. Compounds that inhibit one of the two enzymes generally inhibit the other as well, and the pair is therefore treated throughout this report as a single pharmacological target rather than as two.

Neither enzyme is restricted to a tissue, and the difference between them shows up one level down, in which cell types express each most strongly.

UniProt (SwissProt), NCBI Gene, the Protein Data Bank in Europe, Ensembl Compara, ClinVar, gnomAD v4, Open Targets Platform and the Human Protein Atlas.

Establishing that a protein pair merits pharmacological inhibition is not the same as establishing that a molecule can achieve it, and this section therefore examines the tractability of the enzyme, setting out the solved structures from which a medicinal chemist would work, the patent families already filed against them, and the compounds that other organizations have advanced into clinical trials.

Of these, the patent position alters the picture most, because a field that carries no trials may nevertheless be one in which several organizations are working: composition-of-matter filings become visible years before a trial does.

KAT6A has 21 solved structures in all, by X-ray diffraction and Solution NMR, of which eight are listed here. KAT6B has three solved structures listed here. A dash in the resolution column is a structure solved by a method that does not report one. A dash in the last column is a structure for which no share of the protein was recorded.

The patent field

Six separate holders have filings against the target, which taken together make the patent field around it highly active. Freedom to operate is the right to make and sell a compound without infringing a patent held by another party, and it is determined by the chemical series on which a program is built rather than by the protein at which that program is directed.

A full-text search of the patent corpus on FreePatentsOnline returns 1,008 patent publications that mention KAT6A, and every publication number in the table was then opened on its own record in the PubChem patent view before it was written down, both retrieved 2026-08-15. Menarini appears in no row because a search of the same corpus returns no filing against the target in its name, a result read against a control search that returns 547 patent publications for that name alone; the compound it has taken into the clinic was originated elsewhere, and the composition-of-matter filing behind it is recorded here under the originator.

Trial registries and the published literature were searched for any record of a KAT6 inhibitor administered to a human subject, and the search returned a densely populated oncology field together with no record of any administration outside oncology.

That asymmetry is the single most consequential finding in this assessment, since it establishes both that the mechanism is no longer speculative in humans and that every claim made about aging rests, ultimately, on evidence collected in patients with metastatic cancer.

Grouped by the stage each program has reached to date, with the square marking the program that has reached the furthest stage; information about a program beyond its registry record is printed under the setting. The sponsor column prints the lead sponsor each registry record names, with the corporate suffix dropped, and a trading name or a parent group is given in the note beneath the cell rather than in place of it; for the two programs with no registry record the name is the one their own disclosures use.

The one trial with published numbers

The results set out below are drawn from NCT04606446, the dose-escalation and expansion study published as a Phase 1 trial and now carried in the registry at Phase 2, reported in Mukohara et al., Nature Medicine 2024, which remains the only study among the programs listed above in which one of these compounds has been administered to patients and the findings published in a peer-reviewed journal.

The response rate has been reported twice, at 30.2 percent in the journal paper and at 37.2 percent in a later conference abstract, and both figures are given here rather than the higher one alone.

The expression ranking against the clinical record

The indication ranking derived from the differential-expression analysis is taken entry by entry below and set against whatever clinical evidence exists for each disease.

The ranking figure is the average of the two genes’ fold changes with the direction set aside, so that a disease in which both genes fall can rank alongside one in which both rise, and the direction is therefore given in words as well; the blood cancers and pulmonary fibrosis are the rows where this matters, because in those diseases both genes are lower rather than higher. P-values are printed as the ranking table records them, except in the two rows for which only a two-significant-figure value is available, so 2.7 × 10⁻⁷ and 2.73 × 10⁻⁷ are the same measurement.

Expression of KAT6A and KAT6B in diseased tissue was compared against matched normal tissue across 139 diseases, and the diseases were then sorted according to whether the two genes move together; a positive log2 fold change indicates that the gene is more highly expressed in the diseased tissue.

That sorting matters more than any single value, because where both genes are raised the direction of effect of blocking the enzyme is obvious, whereas where both are lowered it is not, and the case for intervening then depends on reading the reduced expression as compensation rather than as a reason to leave the pathway alone. That inference is the weakest load-bearing step in the entire argument, and the conclusions drawn from it should be read with that in mind.

Fold changes are the PandaOmics differential expression readings for diseased tissue against matched normal tissue, which is the only comparison behind every number in this section. Rows are sorted by tier and then by rank within the tier. The bar under each fold change draws that number on one scale shared by both genes: it runs right from the zero line where the gene is raised in diseased tissue and left where it is lowered, and reaches full length at half a log2 unit. Significance values are recorded for the first tier only, so a dash marks a value that is not available rather than a result that missed a threshold. No sample size, contributing dataset or statistical test is recorded behind any of them, so they can be read as the source’s ordering and not as a result reproducible from what is printed. Click a column heading to sort.

Both genes were additionally scored against roughly 1,000 indications using 23 measures drawn from disease datasets, from the published literature and from grant funding. Language-model scores were excluded from the composite, and umbrella terms such as cancer and sarcoma were removed so that they could not displace specific diseases from the upper positions of the ranking.

This ranking addresses a different matter from the expression comparison, in that expression records the direction in which the genes move in diseased tissue, whereas the ranking records the weight of accumulated evidence connecting a gene to a given disease. A disease may therefore score highly on one measure and poorly on the other, and the age-related indications show precisely that pattern.

A dash means the gene did not rank the indication at all.

The full list runs to twelve; the eight highest-ranked are shown. Indication names are printed in lower case here and in sentence case in the table above, so a disease appearing in both tables is the same disease under two capitalizations.

The composite ranking that drives much of this report carries its own reliability appraisal, which is reproduced here in full: it grades the evidence behind each position, identifies the measures on which weight should not be placed, and names the databases from which each figure was drawn.

That appraisal constitutes a limit on the sections above rather than a footnote to them, since a ranking whose weaker components are explicitly labeled low-confidence can be read critically, and a reader assessing the target is entitled to the same view of its reliability as was available when the ranking was constructed.

Two of the ten grades are accompanied by a stated basis, whereas the rest are recorded without one, and a dash marks the position that a stated basis would otherwise occupy.

Only one of the fifteen retrievals records the date on which it was made, and because the remainder are undated, a value that has since changed in the source database cannot be distinguished from one that has not.

A rank is a position in a list, and a position is not in itself a reason, so this section sets out the mechanistic case for the involvement of the enzyme in the diseases placed at the top of the ranking, together with the experiments that would establish whether that involvement is causal.

The reasoning is given in full rather than summarized, because it is the part of the analysis with which a reader can properly argue, and paraphrase would place this report between the reader and the argument.

Five mechanistic arguments appear below alongside five proposed next steps; because the indications are not named identically in the two lists, one of the headings appears word for word in the other list, and the two sets are therefore presented unpaired rather than aligned row by row.

None of the arguments in the right-hand column has been checked against the papers it cites.

Every one of these has a recorded result that would rule it out.

The case for treating KAT6 as a target in aging biology rests on two propositions: first, that both genes lie on pathways already recognized as drivers of aging, and second, that their inhibition drives damaged cells into a permanent growth arrest without compromising the viability of neighboring healthy cells.

The first proposition is well supported and largely uncontested, whereas the second carries the weight of the argument and faces an objection that must be answered before it can be accepted, namely that the arrest it seeks to induce is the same cellular state that other aging programs are devoting substantial resources to eliminating.

Confidence is a graded judgment on a five-point scale rather than a measured quantity.

The five arguments are numbered, and they are set out here in that order.

Two bodies of data carry the aging argument, and the first of them is drawn from age predictors and age-associated models, most of which read methylation at particular sites, and which between them carry KAT6A or KAT6B among their inputs in the entries counted below; the second is drawn from donated tissue of people who were not ill, in which the expression of both genes changes with the age of the donor.

Both of these are associations, and neither demonstrates that changing the enzyme changes the aging process; no published work links a KAT6 inhibitor to an epigenetic clock reading at all.

Coefficients, ranks and readings are transcribed from the integrated table in which they were first assembled, except where a row states that this report reads the result otherwise. The clocks do not share a sign convention, because the age-shift model turns its classifier’s log-odds into an age score by multiplying through a negative scaling factor released with the model, so a negative coefficient there means older where in a chronological-age model a positive one does. The entry and clock totals given here are taken from the two per-gene sections of that table. Half of the eight denominators are this report’s rather than the review’s: the mammalian predictors are ranked against the number of methylation sites each model keeps rather than the number of genes those sites fall in, and the transcriptomic models against the count of genes in the vector their authors released. The other four are the review’s, carried as it gave them, and this report has not re-derived them.

Only the gene collection here is curated by hand; the remaining rows are coverage checks against a trial registry, a compound database and the published literature. All are revised over time, so the table records the state of each at the moment it was checked.

None of these compounds is selective for KAT6, since the four named natural products inhibit several acetyltransferases and much else besides, which is why they stand as precedent rather than as evidence about this target.

Having established the case for placing the pair within aging biology, this section sets out the design such a program would take and the risks it should expect to encounter, given in full because the design remains informative whether or not a reader accepts the argument that precedes it.

The severity grades below are those carried in the risk table itself, with the single exception recorded in the caption beneath it, and the reasoning set out in the final row is likewise the review’s own rather than this report’s.

Recommended sequence of work

The steps below set out the sequence as it would be run, with an indicative cost against each step.

1. Instrument the oncology trials that are already running.

   GrimAge2 and PhenoAge clocks, p16 and secretory-phenotype panels, GDF15, and H3K23 acetylation in blood cells would be added to studies that are enrolling now. This is the highest-value and lowest-cost action available, because it generates the aging readouts that do not yet exist without requiring a new trial.

   Cost: low.
2. Run a low-dose paradigm study in aged animals.

   Aged mice would be dosed intermittently at a level below the oncology exposure, with senescence burden, secretory phenotype, immune composition and function measured. This study settles the argument over direction, which is the unresolved point that blocks everything downstream.

   Cost: medium.
3. Frame the near-term program as cancer interception.

   The program would target high-risk populations inside the oncology indication rather than present a healthy-aging claim. This framing keeps the regulatory path and the endpoint definition on ground that already exists.

   Cost: low.
4. Differentiate the molecule from the compounds already in the clinic.

   The program would pursue selectivity between the two paralogs, or a degrader, so that the chronic-use profile is not the one Pfizer already has. Several clinical-stage programs address the same mechanism at the same dose intensity.

   Cost: high.

Orthogonal analyses not yet run

One of these is already part-done: the DepMap dependency read named in the fourth item has been carried out, and its result, that neither gene is common-essential in cancer cell lines, is load-bearing elsewhere in this report. The remaining five are open, and they appear in the order the target evaluation itself gives them, which that report sets by cost rather than by importance, although it attaches no cost figure to any of them. The first three come cheapest in the target evaluation’s own ordering, and the first two are the ones whose results could contradict the indication list rather than merely qualify it, so all three should be run before an expansion indication is chosen.

1. Run a phenome-wide association study on loss-of-function and coding variants in UK Biobank or FinnGen.

   Heterozygous carriers of loss-of-function variants provide the closest available approximation to lifelong pharmacological inhibition. Such carriers are close to absent for both genes, at nine observed loss-of-function variants against nearly two hundred expected for one and ten against a comparable expectation for the other, so in practice the coding variants would have to carry the study. The reading also holds for KAT6A more cleanly than for KAT6B, whose truncating transcripts escape nonsense-mediated decay and leave a shortened protein behind rather than simply less of the normal one. A carrier phenotype matching one of the proposed indications would support that indication; no phenotype at all would weaken every genetic argument advanced in this report.

2. Run drug-target Mendelian randomization with colocalization and a bidirectional test.

   This separates a gene that drives a disease from one that merely responds to it. The indication ranking cannot make that separation, because it is a composite of 23 metrics drawn from omics datasets, literature volume and grant funding, so a high rank can be produced by how much has been written about a disease as readily as by how the gene behaves in it. That distinction is the one on which the entire indication list ultimately rests.

3. Measure expression by cell type in a single-cell atlas of the diseased tissue.

   Expression measured in whole tissue also rises when the cells carrying the gene become more numerous. This says whether the gene is genuinely more active in a given cell, or whether the tissue simply has more of that cell in it.

4. Read the DepMap dependency scores for the cancer indications, and the published mouse knockout phenotypes, which the International Mouse Phenotyping Consortium does not hold for either gene.

   A knockout phenotype in the relevant organ would lend independent support to the indication. A DepMap dependency would say whether tumor cells need the gene in order to survive, which is the premise on which an oncology program depends.

5. Run an on-target safety scan across healthy tissues.

   Both genes cause developmental syndromes when disrupted, so the identity of the healthy tissues an inhibitor would affect remains undetermined rather than a formality.

6. Run a senescence and secretory-phenotype panel in aged animals, for the aging indications.

   No published experiment has tested whether changing either gene changes lifespan or senescent burden. This would be the first result bearing directly on the aging case, rather than by inference from expression.

None of these constitutes a decision in itself; each is a piece of work whose result would change the conclusions of this report, which is the only useful criterion by which the cost of performing it can be judged.

The recommendation above is bounded by the measurements that have not been made, and this part collects those bounds in one place: first the evidence that would have to exist before an aging indication could be regarded as supported, and then the reliability that the composite ranking driving much of this report claims for itself.

The evidence assembled in this assessment supports development of a KAT6 inhibitor in oncology and does not, on present data, support a dedicated aging program. That verdict does not rest on an absence of evidence alone, because two published studies, one in aged bone-marrow stromal cells and one in hematopoietic stem cells, report that reducing KAT6A or KAT6B activity moves an aged tissue further in the direction of age rather than away from it. The experiments set out above would test that question directly, although neither they nor the remainder of the program described would establish causation or characterize the consequences of chronic exposure, and both of those gaps would have to close before the position taken here could be revised.

KAT6A and KAT6B satisfy the conventional criteria for pharmacological tractability, in that selective small-molecule inhibitors of the acetyltransferase domain the two enzymes share have been synthesized and are protected by composition-of-matter filings held by several organizations, and target engagement is demonstrable in treated patients, in whom the histone mark the domain deposits is suppressed at tolerated doses. Structural coverage is the uneven part of that case, because the structures deposited for KAT6B resolve the modules by which it reads histones and binds DNA rather than the active site itself. Tractability is nonetheless not among the questions this assessment leaves unresolved, and the gap in the structural record does not put it back among them.

Demonstrated clinical benefit, by contrast, remains confined to oncology, since every asset retrieved, every dosed participant and every efficacy read-out identified in the course of this assessment falls within that field. The peer-reviewed response data come from patients with estrogen receptor-positive, HER2-negative metastatic breast cancer, and the one further program to have reported tumor responses disclosed them through a company announcement rather than a journal, so the two read-outs do not carry equal weight. A program directed at aging would in either case begin without clinical observation specific to the indication it proposes to treat.

The rationale for an aging indication and the principal risk attached to it derive, on present evidence, from a single observation, in that the induction of a permanent growth arrest is the mechanism by which the inhibitor is thought to arrest a tumor, whereas the elimination of cells already in that state is the objective pursued across much of the longevity field. No experiment reported to date establishes which of the two effects dominates in a healthy older animal, and until one does, the evidence that most strongly supports the target constitutes the principal argument against this particular application of it.

Primary literature

The publications below are the peer-reviewed literature this assessment rests on, and every entry carries both its PubMed identifier and its digital object identifier, so that each can be checked against the primary record rather than taken on trust. The list is not a complete inventory of what the page draws on: clinical trials are cited by registration number where they appear, and the higher of the two response rates reported in the clinical section rests on a conference abstract that carries no identifier of either kind.

1. Baell JB et al. (2018) Inhibitors of histone acetyltransferases KAT6A/B induce senescence and arrest tumour growth NaturePMID 30069049 doi:10.1038/s41586-018-0387-5
2. Carapeti M et al. (1998) A novel fusion between MOZ and the nuclear receptor coactivator TIF2 in acute myeloid leukemia BloodPMID 9558366 doi:10.1182/blood.v91.9.3127
3. Collins HM et al. (2006) MOZ-TIF2 alters cofactor recruitment and histone modification at the RARbeta2 promoter: differential effects of MOZ fusion proteins on CBP- and MOZ-dependent activators J Biol ChemPMID 16613851 doi:10.1074/jbc.M602633200
4. Fei D et al. (2021) KAT6A regulates stemness of aging bone marrow-derived mesenchymal stem cells through Nrf2/ARE signaling pathway Stem Cell Res TherPMID 33541408 doi:10.1186/s13287-021-02164-5
5. Huang B et al. (2020) Inhibition of histone acetyltransferase GCN5 extends lifespan in both yeast and human cell lines Aging CellPMID 32157780 doi:10.1111/acel.13129
6. Huang F et al. (2016) Regulation of KAT6 acetyltransferases and their roles in cell cycle progression, stem cell maintenance, and human disease Mol Cell BiolPMID 27185879 doi:10.1128/MCB.00055-16
7. Katsumoto T et al. (2006) MOZ is essential for maintenance of hematopoietic stem cells Genes DevPMID 16702405 doi:10.1101/gad.1393106
8. Katsumoto T et al. (2022) MOZ is critical for the development of MOZ/MLL fusion-induced leukemia through regulation of Hoxa9/Meis1 expression Blood AdvPMID 35947126 doi:10.1182/bloodadvances.2020003490
9. Khokhar ES et al. (2020) Aging-associated decrease in the histone acetyltransferase KAT6B is linked to altered hematopoietic stem cell differentiation Exp HematolPMID 32014431 doi:10.1016/j.exphem.2020.01.014
10. Kindle KB et al. (2005) MOZ-TIF2 inhibits transcription by nuclear receptors and p53 by impairment of CBP function Mol Cell BiolPMID 15657427 doi:10.1128/MCB.25.3.988-1002.2005
11. Kirfel P et al. (2020) Inhibition of histone acetylation and deacetylation enzymes affects longevity, development, and fecundity in the pea aphid (Acyrthosiphon pisum) Arch Insect Biochem PhysiolPMID 31498475 doi:10.1002/arch.21614
12. Largeot A et al. (2016) Expression of the MOZ-TIF2 oncoprotein in mice represses senescence Exp HematolPMID 26854485 doi:10.1016/j.exphem.2015.12.006
13. Miyamoto R et al. (2020) Activation of CpG-rich promoters mediated by MLL drives MOZ-rearranged leukemia Cell RepPMID 32997997 doi:10.1016/j.celrep.2020.108200
14. Morselli E et al. (2009) Autophagy mediates pharmacological lifespan extension by spermidine and resveratrol Aging (Albany NY)PMID 20157579 doi:10.18632/aging.100110
15. Mousavi N & Yang XJ (2025) Lysine Acetyltransferase 6 Complexes in Neurodevelopmental Disorders and Different Types of Cancer Results Probl Cell DifferPMID 40593218 doi:10.1007/978-3-031-91459-1_14
16. Mukohara T et al. (2024) Inhibition of lysine acetyltransferase KAT6 in ER+HER2− metastatic breast cancer: a phase 1 trial Nat MedPMID 38824244 doi:10.1038/s41591-024-03060-0
17. Rokudai S et al. (2013) MOZ increases p53 acetylation and premature senescence through its complex formation with PML Proc Natl Acad Sci USAPMID 23431171 doi:10.1073/pnas.1300490110
18. Sheridan M et al. (2024) The small inhibitor WM-1119 effectively targets KAT6A-rearranged AML, but not KMT2A-rearranged AML, despite shared KAT6 genetic dependency J Hematol OncolPMID 39380002 doi:10.1186/s13045-024-01610-0
19. Smolko AE et al. (2024) A MOZ-TIF2 leukemia mouse model displays KAT6-dependent H3K23 propionylation and overexpression of a set of active developmental genes Proc Natl Acad Sci USAPMID 38889153 doi:10.1073/pnas.2405905121
20. Tan Y et al. (2025) Lysine Acetyltransferase 6 in Health and Disease MedCommPMID 41357671 doi:10.1002/mco2.70520
21. Xi Z et al. (2026) Dancing with KAT6A: current advances and therapeutic potential in oncology of KAT6A inhibitors Bioorg ChemPMID 41604973 doi:10.1016/j.bioorg.2026.109533
22. Yang XJ & Ullah M (2007) MOZ and MORF, two large MYSTic HATs in normal and cancer stem cells OncogenePMID 17694082 doi:10.1038/sj.onc.1210609

Terms used here

Histone acetyltransferase

An enzyme that attaches a small chemical tag to the proteins DNA is wound around, which loosens the packing and lets genes be read. KAT6A and KAT6B are two of them, from the group numbered KAT5 to KAT8 known as the MYST family. Also written as: HAT, KAT, MYST

Log2 fold change

A measure of the difference in a gene’s activity between diseased and healthy tissue, on a doubling scale. A value of 1 means twice as much; 0.4 means about a third more. Also written as: log2FC, logFC

Epigenetic clock

A statistical model that estimates an age, or a lifespan, from a molecular measurement. Some of these models read chemical marks on DNA, whereas others read how strongly genes are transcribed, and several of those drawn on here were fitted across mammalian species rather than within our own. The earliest were trained against the calendar; later ones were trained against health and survival, so the years they return estimate condition rather than time elapsed. A gene being one of the inputs says the model uses it, and not that the gene causes aging.

Cellular senescence

A state in which a cell stops dividing permanently but stays alive and keeps signaling to its neighbors, and the state into which blocking KAT6 drives cells.

Secretory phenotype

The set of inflammatory signals a senescent cell releases. It is the reason senescent cells are thought to damage the tissue around them. Also written as: SASP

Tolerance of losing one gene copy

A measure of how badly a species tolerates losing one working copy of a gene. Low values mean the gene is rarely lost in healthy people, so inhibiting it warrants caution. Also written as: LOEUF

Genome-wide association study

A scan across the genomes of many people for common inherited differences that occur more often in those with a trait. It shows a stretch of chromosome is involved, not which gene in it does the work. Also written as: GWAS

Mendelian randomization

A method that uses inherited genetic differences to test whether changing a gene’s activity actually changes a disease, rather than merely tracking with it.

Estrogen-receptor-positive, HER2-negative breast cancer

The tumor cells carry the receptor that estrogen acts on, and do not overproduce the HER2 growth receptor. It is the setting of the most advanced KAT6 program, and fulvestrant, the drug paired with the inhibitor in every one of that program’s patient trials, acts on that same receptor. Also written as: ER+/HER2−, ER+

Blinded independent central review

Scans in a trial are read by assessors who do not know which treatment a patient received, so the assessment cannot drift toward the outcome the sponsor expects. Also written as: BICR

Patient-derived xenograft

A fragment of a patient’s tumor grown in a mouse, used to test a drug against that tumor before it is administered to patients. Also written as: PDX

Arboleda-Tham syndrome

The developmental disorder caused by a damaging change in one of the two copies of KAT6A. Losing a working copy outright is one route to it; missense and splicing variants that leave an altered protein behind are others, and the variants that truncate the protein latest are associated with the more severe presentation. It is also written as KAT6A syndrome, and the two names denote one condition. Also written as: KAT6A syndrome

Neutropenia

A shortage of the white blood cells that fight bacterial infection. Grade 3 and grade 4 mean severe and life-threatening.

Dysgeusia

A distorted sense of taste. It was the most frequent treatment-related event in the Phase 1 trial and was reported only at grades one and two. No patient discontinued treatment because of it, and at the recommended dose it prompted no change of dose either, though during dose escalation a few patients reduced their dose on account of it. Whether a taste disturbance this common would be tolerated across years of dosing, rather than across the months an oncology trial runs, has not been tested.

Reference number of a published paper

The number the United States National Library of Medicine assigns to each publication it indexes, called a PubMed identifier. Every entry in the reference list below carries one, alongside a digital object identifier, so that a reader can retrieve the primary record rather than take the citation on trust. Not every source this assessment rests on is a publication that PubMed indexes, and the ones that are not are attributed at the point where they are used. Also written as: PMID